Amphenol's $10 Billion Quarter Comes With a Slowing Datacom Ramp

Record orders, record sales, record margins — and a stock that has already had its re-rating, flat over 90 days and 8.6% below its October high

APH · Earnings Call · 2026-10-28

A scale event, not just a beat

Amphenol's third quarter was not simply a beat. Sales reached a record $10.2 billion, up 65% year over year and 37% organically, with adjusted diluted EPS of $0.80, up 74%. “Third quarter sales were up 65% ... and 37% organically compared to the third quarter of 25.” — Craig A. Lampo, Chief Financial Officer (CFO) · 2026-10-28 Orders hit $12.17 billion, nearly double the prior-year quarter, for a 1.19 book-to-bill. For a $199.6 billion market-cap components company, the more important change was inside the mix. IT Datacom became 48% of sales. “The IT Datacom market represented 48% of our sales in the third quarter. Sales in this important market grew by a very strong 100% in U.S. dollars and 77% organically.” — Richard Adam Norwitt, Chairman and CEO · 2026-10-28 AI is now roughly three-quarters of that business. The electronic industry is no longer a diversified end-market mosaic for Amphenol; it is an AI interconnect ramp with an industrial and defense tail. The prior quarter's filed numbers already showed the torque: revenue of $8.8 billion, up 55% year over year, with operating margin at 29.5%, up 4.4 percentage points. Q3 extended it: 29.7% adjusted operating margin, up 22 basis points year over year despite the dilution from recent acquisitions. The old Amphenol was a 20%-margin connector compounder. The new one is approaching 30% and debating internally whether 29.7% is a floor.

The order book is the tell — and the warning

The cleanest evidence that demand is not mere replenishment: orders essentially doubled. The CEO pushed back on neocloud funding worries, saying “the vast majority of the demand that I think we are seeing, that our customers are seeing, are from some of the world's most highly capitalized companies” — Richard Adam Norwitt, Chairman and CEO · 2026-10-28. That is a direct answer to the quality-of-spending worry hanging over AI infrastructure. But the "open order aperture" explanation is not new. In January: “we have seen customers open up their order window for in particular related to significant plans that they have of investments related to AI” — Adam Norwitt, Chief Executive Officer · 2026-01-28. In July: “They need more of everything. More high-speed copper. More fiber optic solutions, and more power solutions.” — Richard Adam Norwitt, Chief Executive Officer (CEO) · 2026-07-29 So the record book is a continuation of an existing behavior, scaled — not a fresh demand signal, and the market should not underwrite it as one. What is genuinely newer is the deceleration implied in guidance. The Datacom business grew 30% sequentially against a mid-teens expectation; Q4 is guided up only mid-single digits from that very high level. The CEO insists it is not supply: “I do not think it is gated to say that it is gated by demand or our abilities. I mean, we have been executing really, really well.” — Richard Adam Norwitt, Chairman and CEO · 2026-10-28 Fine — but at 48% of sales it is still the first explicit sequential slowing in the AI ramp, and that is what the tape is sniffing at.

The tape cooled while the P&L accelerated

The global price tape is no longer voting for the theme with the same enthusiasm. AI data center sits among the top 360-day advancers at roughly +53%, but over the last 90 days the same cluster is down about 5%, with 50 decliners against 6 advancers; data center capacity and high-performance compute are negative over both 90- and 30-day windows. Amphenol's own tape agrees: the full history is up 2,781% since 2010, but the last 90 days are essentially flat at +2.0%, and the shares sit 8.6% below the October 6 peak of $88.62. Sell-side price targets were raised in early October — just before that peak. The quarter is excellent; the market has already moved on to asking what happens next. The next content leg is power. Management has been building the case that data centers need more interconnect across high-speed copper, optics and power. Power interconnect is the least appreciated of the three, and it becomes more important as racks move to high voltage designs such as 800-volt architectures. The framing is simple:

these data centers are basically trying to do a simple thing, which is turn electrons into tokens. And if we can help our customers use less of those electrons through the efficiency of the power interconnect while keeping everything safe and secure...
Richard Adam Norwitt, Chairman and CEO · 2026-10-28

That is a good pitch, but also a multi-year design-in story, not a fourth-quarter earnings lever. The same holds for the fiber optic breadth added with CommScope, which management now expects to more than double in IT Datacom this year.

Concentration, and the margin and balance-sheet buffer

One analyst asked whether 48% IT Datacom forces a rethink of the portfolio. The answer was revealing: “if you take out, by the way, AI, and the growth in AI, we are driving organic growth in the high teens percentage for the company” — Richard Adam Norwitt, Chairman and CEO · 2026-10-28. That is a strong non-AI number — but it also concedes the point. Amphenol is now managed around an AI cycle, with "one foot on the gas and one on the brake," as the CEO put it, via acquisitions: Empire, Advanced PCB, DexScreen and Humanetics. Together they are small against $37 billion of guided sales, but they broaden defense, fiber and sensing. The defense business is already growing 42% with what management frames as a multiyear replenishment cycle — the best non-AI hedge in the portfolio. Margins are the buffer. A 29.7% adjusted operating margin is remarkable for a components company, and the CFO argues it is not the ceiling. The swing factor is CommScope, running in the low-20s operating margin versus the corporate average; closing that gap adds another layer of EPS leverage. The balance sheet paid for it: effective net cash of -$11.8 billion and liabilities to assets at 65.1%, up 10.3 percentage points year over year. Net leverage of 1.1x means this is not a solvency story — but the ramp is now being funded with a more levered balance sheet, and that is the price of the concentration.

Bottom line

The quarter proves AI interconnect demand is real, and that Amphenol is taking more than its fair share of the rising intensity of content per system. But the easy re-rating looks done. The stock is flat over 90 days and 8.6% below its high, the AI data center basket has cooled, and the Q4 guide implies the first sequential slowdown in the datacom ramp. The story has shifted from 'is AI demand real?' to 'how much AI concentration is tolerable?' Amphenol's answer is more interconnect, in more formats, across more customers. The tape's answer, for now, is a shrug. That tension — not the headline beat — is what makes this report worth watching.